[Bug target/40983] New: The scheduler incorrectly swaps MMX and floating point instructions

Hi

This example fails, because in function "f", the scheduler incorrectly swapped
floating point store to "c" and load of mmx registers.

Compile with -O2 -march=pentium-mmx


--
           Summary: The scheduler incorrectly swaps MMX and floating point
                    instructions
           Product: gcc
           Version: 4.4.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: mikulas at artax dot karlin dot mff dot cuni dot cz
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40983

------- Comment #1 from mikulas at artax dot karlin dot mff dot cuni dot cz  2009-08-06 04:12 -------
Created an attachment (id=18310)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18310&action=view)
A failing example


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40983

------- Comment #2 from mikulas at artax dot karlin dot mff dot cuni dot cz  2009-08-06 04:15 -------
Assembler output:

f:
        pushl   %ebp
        movl    %esp, %ebp
        subl    $16, %esp
        movq    %mm0, -8(%ebp)
        movq    %mm1, -16(%ebp)
        emms
        fldl    a
        faddl   b
        movq    -8(%ebp), %mm0
        paddd   -16(%ebp), %mm0
        fstpl   c
        leave
        ret


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40983

------- Comment #3 from ubizjak at gmail dot com  2009-08-06 10:47 -------
Hm, we can fix this by teaching scheduler that every access to %mm registers
clobbers FP state. Since scheduler already depend all x87 instructions on
Top-Of-Stack (TOS) register, we can perhaps extend this requirement for all
instructions that use %mm reg.

Something like the proof-of-concept patch bellow:

--cut here--
Index: sched-deps.c
===================================================================
--- sched-deps.c        (revision 150503)
+++ sched-deps.c        (working copy)
@@ -1898,7 +1898,12 @@
            sched_analyze_reg (deps, FIRST_STACK_REG, mode, code, insn);
          sched_analyze_reg (deps, FIRST_STACK_REG, mode, USE, insn);
        }
+
+#ifdef FIRST_MMX_REG
+      if (regno >= FIRST_MMX_REG && regno <= LAST_MMX_REG)
+       sched_analyze_reg (deps, FIRST_STACK_REG, VOIDmode, CLOBBER, insn);
 #endif
+#endif
     }
   else if (MEM_P (dest))
     {
@@ -2044,7 +2049,12 @@
            sched_analyze_reg (deps, FIRST_STACK_REG, mode, USE, insn);
          sched_analyze_reg (deps, FIRST_STACK_REG, mode, SET, insn);
        }
+
+#ifdef FIRST_MMX_REG
+      if (regno >= FIRST_MMX_REG && regno <= LAST_MMX_REG)
+       sched_analyze_reg (deps, FIRST_STACK_REG, VOIDmode, CLOBBER, insn);
 #endif
+#endif

        if (cslr_p && sched_deps_info->finish_rhs)
          sched_deps_info->finish_rhs ();
--cut here--

Otherwise, you can just add "asm volatile ("")" as a scheduling barrier.


--

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-08-06 10:47:08
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40983

------- Comment #4 from mikulas at artax dot karlin dot mff dot cuni dot cz  2009-08-23 19:18 -------
The patch works fine.


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40983

------- Comment #5 from sezeroz at gmail dot com  2009-10-16 19:45 -------
Any progress on this?


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40983

------- Comment #6 from mikulas at artax dot karlin dot mff dot cuni dot cz  2009-11-11 23:06 -------
I think you can commit it to gcc. I don't see a reason why it shouldn't be
committed.


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40983